NIST Validation Studies on the

3500 Genetic Analyzer

Erica ButtsNational Institute of Standards and Technology

Applied Biosystems HID UniversityAtlanta, GA

July 12, 2011

htt

ps:/

/pro

du

cts

.ap

plie

db

iosyste

ms.c

om

/ab

/en

/US

/htd

ocs/p

rod

uctM

gr/

ima

ge

s/3

50

0_

fam

ily_

big

.jp

g

Outline

• Details of the new ABI 3500

• Validation design and results with Identifiler and Identifiler Plus – Precision

– Injection parameters and reaction setup

– Setting Analytical Thresholds

– Setting Stochastic Thresholds

• Overview of signal normalization

Details of the new ABI 3500

Improved sealing for better

temperature control

Better seal around

the detector

No lower pump block

(less polymer waste)

Reagents prepackaged

with RFID tags

Primary Differences Between 31xx

and 3500

31xx Instruments

• Argon ion (Ar+) lasers with

488/514 nm wavelengths for

fluorescence excitation

• 220V power requirement

• Optimal signal intensity 1500-

3000 RFU

• Currently validated and

operational in most forensic

laboratories

3500 Instruments

• Single-line 505 nm, solid-state long-life laser

• Smaller footprint

• 110V power requirement

• Optimal signal intensity can approach 20,000-30,000 RFU

• Normalization of instrument-to-instrument signal variability– Ability to increase or decrease

overall signal

• Requires the use of GeneMapper IDX v1.2

What is Validation?

Section 1.1 (SWGDAM Revised Validation Guidelines) Validation is the process by which the scientific community acquires the necessary information to:

(a) Assess the ability of a procedure to obtain reliable results.

(b) Determine the conditions under which such results can be obtained.

(c) Define the limitations of the procedure.

The validation process identifies aspects of a procedure that are critical and must be carefully controlled and monitored.

Reliability, Reproducibility, Robustness

Aspects of Validation

• Reliability– Size Standard Comparison

• What is the difference between LIZ 500 and LIZ 600 v2.0?

– Injection Parameters• What are the best injection parameters for typable data?

• Adjustment alongside PCR reaction volume?

• Reproducibility– Precision

• Is the precision comparable to the 3130xl?

– Concordance• Are the correct allele calls being made?

• Robustness– Sensitivity

• How do the thresholds affect the analysis of data?

– Mixtures (current work in progress)• How often is the minor component identified?

Experimental SummaryTest Types of Samples Used

Number

Examined

Size Standard

Comparison

16 Allelic Ladders per size standard

(LIZ 500 vs. LIZ 600 v2.0)32

Injection

Parameters

3 samples heterozygous at all loci including

Amelogenin 1 ng DNA input

3 samples/injection

time

Total 15 samples

Precision

Allelic Ladders 24

3 samples heterozygous at all loci including

Amelogenin6

SensitivityDilution series of 3 samples heterozygous at all loci

including Amelogenin

4 Replicates of

each dilution

series

Total 84 samples

Concordance50 genomic DNA samples

60SRM 2391b: 10 genomic DNA samples

MixturesMixture dilution series of 2 samples heterozygous at

all loci including Amelogenin28

Total Number of Samples 249

Identical experiments for Identifiler and Identifiler Plus

Reaction Setup

Identifiler

• PCR Volume: 12.5 µL

– Primer Mix

– Master Mix

– Taq Gold Polymerase

• 28 cycles

• 1 ng DNA target input

unless otherwise stated

Identifiler Plus

• PCR Volume: 12.5 µL

– Primer Mix

– Master Mix

– No separate Taq/enzyme added

• 28 cycles

• 1 ng DNA target input unless

otherwise stated

Identifiler Plus is optimized to overcome inhibition with an improved buffer mix

Cleaner baseline and improved heterozygote peak balance

Sample Preparation for Capillary

Electrophoresis

• 17.4 µL HiDi + 0.6 µL LIZ 600 v2.0 per sample

• 2 µL Sample or Allelic Ladder

• Centrifuge for 1 min to mix

• Aliquot 10 µL into a separate plate

– Centrifuge both plates 1 min

• Plates run on 3130xl and 3500 simultaneously

Size Standard Comparison

1 2

A LIZ 500 LIZ 600 v2.0

B LIZ 600 v2.0 LIZ 500

C LIZ 500 LIZ 600 v2.0

D LIZ 600 v2.0 LIZ 500

E LIZ 500 LIZ 600 v2.0

F LIZ 600 v2.0 LIZ 500

G LIZ 500 LIZ 600 v2.0

H LIZ 600 v2.0 LIZ 500

Individual master mixes created for LIZ 500 and LIZ 600 v2.0 with Identifiler/Identifiler Plusallelic ladders

Injected twice on 3130xl– Standard injection of 3 kV

for 10 seconds

Injected 3 times on 3500– Default Injection of 1.2 kV

for 15 seconds

It is important to determine if one size standard can be used consistently on

both the 3130xl and 3500 for proper comparison

n=16 Allelic Ladders

Size Standard ComparisonLIZ 500

LIZ 600 v2.0

The 250 bp and sometimes 340 bp

peaks do not size reproducibly and

are often not designated as

standard peaks

0.00

0.05

0.10

0.15

0.20

0.25

0.30

100 150 200 250 300 350

SD

Alle

le S

ize

, b

p

Mean Allele Size, bp

LIZ 500

LIZ 600

s(LIZ 500): 0.061 bp

s(LIZ 600): 0.049 bp

3500

s(LIZ 500): 0.061 bp

s(LIZ 600): 0.049 bp

Size Standard Comparisonn=16 Allelic Ladders

LIZ 600 v2.0 generated the most linear results on both the 3130xl and 3500 and was

used as the size standard on both instruments for remaining testing

This upward shift is likely due to the

shifting of the 340 bp peak due to a

fluctuation of temperature across or

between runs

D8S

1179

D21S

11

D7S

820

CS

F1

PO

D3S

1358

TH

01

D13s317

D16S

539

D2S

1338

D19S

433

vW

A

TP

OX

D18S

51

AM

EL

D5S

818

FG

A

0.00

0.05

0.10

0.15

0.20

0.25

0.30

100 150 200 250 300 350

SD

Alle

le S

ize

, b

p

Mean Allele Size, bp

LIZ 500

LIZ 600

Locus Average

s(LIZ 600): 0.049 bp

s(LIZ 500): 0.128 bp

s(LIZ 500): 0.128 bp

s(LIZ 600): 0.049 bp

3130xl

Injection Parameters

• Injection voltage/time:

– 1.2 kV for 15 sec

– 1.2 kV for 10 sec

– 1.2 kV for 7 sec

– 1.2 kV for 5 sec

– 1.2 kV for 3 sec

Standard injection parameters set

based on samples with:1. No pull-up present

2. No drop out present

n=20: Identifiler n=15: Identifiler Plus

Identifiler

Identifiler Plus

~33,000 RFU

~22,000 RFU

~16,000 RFU

~13,000 RFU

~2,600 RFU

15 sec

10 sec

7 sec

5 sec

3 sec

35,000 RFU

Identifiler Plus

Sizing Precision

1 2 3 4

A Identifiler EBIdentifiler

PlusEB

B Neg Identifiler NegIdentifiler

Plus

C Identifiler EBIdentifiler

PlusEB

D Neg Identifiler NegIdentifiler

Plus

E Identifiler EBIdentifiler

PlusEB

F Neg Identifiler NegIdentifiler

Plus

G Identifiler SampleIdentifiler

PlusSample

H Sample Identifiler SampleIdentifiler

Plus

Identifiler and Identifiler Plusallelic ladders in checkerboard pattern

Neg: PCR blank– PCR primers + water

EB: Extraction blank– PCR primers + extraction

eluent

Sample: 1 ng heterozygous sample at 15 loci plus Amelogenin

Injected 3 times with the newly determined injection parameters

Identifiler Identifiler Plus

n=24 Allelic Ladders, n=6 Samples

0.00

0.03

0.06

0.09

0.12

100 150 200 250 300 350

SD

Alle

le S

ize

, b

p

Mean Allele Size, bp

Ladder

Sample

s(Ladder): 0.044 bp

s(Sample): 0.051 bp

0.00

0.03

0.06

0.09

0.12

100 150 200 250 300 350

SD

Alle

le S

ize

, b

p

Mean Allele Size, bp

Ladder

Sample

s(Ladder): 0.044 bp

s(Sample): 0.053 bp

Precision of Base Pair Sizing

No significant difference between 3130xl and 3500

n=24: Allelic Ladders, n=6 Samples

35003130xl

No significant difference between Identifiler and Identifiler Plus

Setting Analytical Thresholds

• Analytical Threshold (AT)– Minimum threshold for data comparison and peak detection in

the DNA typing process

• AT values can be calculated using negative controls– Analyze with threshold set at 1 RFU

– Calculate average RFU noise per dye channel

• AT values can be calculated using a DNA dilution series– Analyze with threshold at 1 RFU

– Remove calls for all alleles and artifacts (stutter, n+4, pull-up, etc)

Butler, J.M. (2009) Fundamentals of Forensic DNA Typing. Elsevier Academic Press: San Diego

Methods For Calculation

• Method 1 (LOD): Average RFU + (3 x Standard

Deviation)

• Method 2: Average RFU + (t1-α,v from student t-table x

Standard Deviation)

• Method 3: 2 x (Ymax – Ymin)

• Method 4 (LOQ): Average RFU + (10 x Standard

Deviation)

Methods 1-3: “Mixture Interpretation: Principles, Protocols, and Practice” workshop at the 21st International Symposium on Human Identification

(San Antonio, TX), October 11, 2010

Calculations Using Negative Controls

Identifiler

Average

RFU Stdev Min RFU Max RFU Method 1 Method 2 Method 3 Method 4

Blue 9 3.3 2 22 19 19 44 42

Green 13 3.6 5 27 24 23 54 49

Yellow 20 4.9 8 31 35 34 62 69

Red 27 7.1 10 50 49 48 100 99

Identifiler Plus

Average

RFUStdev Min RFU Max RFU Method 1 Method 2 Method 3 Method 4

Blue 9 3.1 3 20 18 18 40 39

Green 13 3.4 4 26 23 23 52 47

Yellow 20 5.1 7 37 36 35 74 72

Red 28 7.2 11 54 49 48 108 99

If calculating analytical threshold using negative controls:

Identifiler: 100 RFU

If calculating analytical threshold using negative controls:

Identifiler Plus: 100 RFU

Calculations Using DNA Dilution Series

If calculating analytical threshold using a DNA dilution series

Identifiler: 140 RFU

If calculating analytical threshold using a DNA dilution series

Identifiler Plus: 120 RFU

Identifiler

Average

RFU Stdev Min RFU Max RFU Method 1 Method 2 Method 3 Method 4

Blue 9 8.4 1 66 34 33 132 93

Green 13 11.5 3 84 48 47 168 128

Yellow 22 11.6 4 88 57 56 176 138

Red 28 8.8 10 80 54 53 160 116

Identifiler Plus

Average

RFU Stdev Min RFU Max RFU Method 1 Method 2 Method 3 Method 4

Blue 10 4.6 3 68 23 23 136 55

Green 16 5.6 3 78 33 32 156 72

Yellow 24 7.9 7 63 48 47 126 103

Red 31 8.9 7 81 57 56 162 120

Single Analytical Threshold Summary

Negative Controls Positive Controls

Ide

nti

fier

Ide

nti

file

r

Plu

s

100 RFU

100 RFU

140 RFU

120 RFU

One Threshold vs. Dye Specific Thresholds

• Evaluation of Method 4 of the DNA dilution series data to

determine the statistical difference between dye channel

analytical thresholds

• Calculated statistical difference using a z-test

Not different

Different

Example: Blue to Red = 44.4• If negative: Not statistically different

– Error bars overlap

– One standard analytical threshold can

be applied to all dyes

• If positive: Statistically different

– Error bars do not overlap

– Dye specific analytical thresholds need

to be applied

Dye Specific Thresholds

• Statistically each dye channel is different for

both Identifiler and Identifiler Plus

– Need to be treated independently

Identifiler

Method 4

(RFU)

AT

(RFU)

Blue 93 95

Green 128 130

Yellow 138 140

Red 116 120

Identifiler Plus

Method 4

(RFU)

AT

(RFU)

Blue 55 55

Green 72 75

Yellow 103 105

Red 120 120

Heat Map ExplanationResults broken down by locus

Green = full (correct) type

Yellow = allele dropout

Red = locus dropout

This is an easy way to look at a lot of data at once

A single profile slice

A replicate slice

One Threshold Dye Specific Thresholds

Total of 15.5% of the loci

improved using dye specific

analytical thresholds rather

than one threshold for all dyes

Heat Map Comparison: Identifiler

PHR=30%

Heat Map Comparison: Identifiler PlusOne Threshold Dye Specific Thresholds

Total of 38.6% of the loci

improved using dye specific

analytical thresholds rather

than one threshold for all dyes

PHR=30%

Setting Stochastic Thresholds

Stochastic Threshold (ST)

– Detection level on an instrument (31xx or 3500)

where a potential sister allele of detected peak may

fall below the analytical threshold

– The value above which it is reasonable to assume

that allelic dropout of a sister allele has not occurred

Butler, J.M. (2009) Fundamentals of Forensic DNA Typing. Elsevier Academic Press: San Diego

Setting Stochastic Thresholds

• Dilution series of three heterozygous samples at

15 loci plus Amelogenin to evaluate where drop

out is first observed

• Total DNA input: 1.0 ng, 0.5 ng, 0.25 ng,

0.10 ng, 50 pg, 30 pg,10 pg in 4 replicates

• Determine RFU value of highest surviving false

homozygous peak per dye channel

Stochastic Threshold

Identifiler: 28 cycles

Standard Injection on 3500:

7 sec @ 1.2 kV inj

n=84 Samples

50 pg

30 pg

10 pg

50 pg

30 pg

50 pg

50 pg

30 pg

30 pg

Stochastic Threshold

Identifiler Plus: 28 cycles

Standard Injection on 3500:

5 sec @ 1.2 kV inj

50 pg

30 pg

10 pg

n=84 Samples

50 pg

50 pg

50 pg

30 pg

30 pg

30 pg

Stochastic Thresholds

n=84 Samples

Identifiler

ST

Blue 345

Green 435

Yellow 410

Red 310

Identifiler Plus

ST

Blue 290

Green 385

Yellow 415

Red 265

Identifiler: 28 cycles

Standard Injection on 3500:

7 sec @ 1.2 kV inj

Identifiler Plus: 28 cycles

Standard Injection on 3500:

5 sec @ 1.2 kV inj

Concordance

• 50 unique male samples

• SRM 2391b: 10 genomic DNA Samples

• All 60 samples concordant between 3130xl and

3500

• Total of 1689 alleles examined

1:1

Mixture Experimental Design

1 2 3 4 5 6 7 8

A Ladder Ladder Ladder Ladder Ladder Ladder Ladder Ladder

B 1:1 1:1 1:1 1:1 1:1 1:1 1:1 1:1

C 2:1 2:1 1:2 1:2 2:1 2:1 1:2 1:2

D 3:1 3:1 1:3 1:3 3:1 3:1 1:3 1:3

E 5:1 5:1 1:5 1:5 5:1 5:1 1:5 1:5

F 7:1 7:1 1:7 1:7 7:1 7:1 1:7 1:7

G 9:1 9:1 1:9 1:9 9:1 9:1 1:9 1:9

H 10:1 10:1 1:10 1:10 10:1 10:1 1:10 1:10

Identifiler Identifiler Plus

• 2 samples heterozygous at 15 loci

plus Amelogenin

• Mixture ratios from 1:1 to 1:10 (and

inverse)

• Samples were injected twice

1:7

1:10

Minor component identified correctly in a 1:10 mixture ratio

n=28 Mixtures

What is Normalization and

how does it work?

Normalization of Data

• Recommended to compare signal between instruments

• Motivation mainly for large laboratories with many instruments– Correct for signal variation between instruments

• Can be used with a single instrument– Correct for signal variation between single and

multiple injections

Normalization Definitions

• Normalization Target (NT)

– Requires the use of LIZ 600 v2.0 size standard

– Average peak heights of 11 peaks within LIZ 600 v2.0

selected for peak height consistency across lots

• Applied within data collection software prior to running samples

• Normalization Factor (NF)

– Adjustment needed for individual samples to reach

the Normalization Target value

– Full signal adjustment (baseline, peaks, artifacts, etc)

• Either increase or decrease signal

Normalization Example

Without

Normalization

NF=1.959

Signal increases

almost by a factor of 2

NF= 0.596

Signal decreases

by almost half

With

Normalization

Theoretical Normalization Target: 2000 RFU

LIZ 600 v2.0

Peak Height

Average:

1021 RFU

LIZ 600 v2.0

Peak Height

Average:

3192 RFU

Conclusions

• The 3500 has proven to be reliable, reproducible and robust

– Out of 498 samples between Identifiler and Identifiler Plus only 5

required reinjection

– Precision within about 0.05 base pairs

• Dye specific analytical thresholds result in less allelic and full

locus dropout than applying one analytical threshold to all dyes

• Stochastic thresholds are linked to analytical thresholds

– If the analytical threshold is adjusted, the stochastic threshold should be

reevaluated along with expected peak height ratios

• Minor contributor successfully identified in as low as a 1:10

mixture

Future Work

• Validation of additional kits

• More extensive review of the impact of thresholds on interpretation– Interaction between analytical and stochastic

thresholds alongside peak height ratios

• More extensive review of normalization– Do thresholds change when employing

normalization?

Acknowledgments

Margaret

Kline

Becky

HillKristen Lewis

O’Connor

Pete

ValloneDave

Duewer

Mike

Coble

John

Butler

Forensic DNA Team DNA Biometrics Team

Funding from the National Institute of Justice (NIJ)

through NIST Office of Law Enforcement StandardsFunding from the FBI S&T Branch

through NIST Information Access Division

Data Analysis

Support

Contact Info:

erica.butts@nist.gov

301-975-5107